Journal of Research of the National Bureau of Standards
AbstractIn this paper several programs for computing the inverse of a matrix are compared primarily on the basis of execution time. Accuracy estimates and two programs that use iterative refinement are included. It is shown that for small matrices, improvement procedures are worthwhile but for large matrices, one must be more careful in their use. Two other points are also brought out: the value of multiplying matrices before taking the norm of a product and the need for some kind of an error estimate to be included in the output of every program.
CiteJ. Res. Natl. Bur. Stand., Sec. B: Math. Sci., Vol. 78B, No. 1, p. 15
OcrABBYY FineReader 8.0
RightsThe Journal of Research of the National Institute of Standards and Technology is a publication of the U.S. Government. The papers are in the public domain and are not subject to copyright in the United States. However, please pay special attention to the